[erlang-questions] Parallel Shootout & a style question

Isaac Gouy <>
Thu Sep 4 17:03:00 CEST 2008


--- "Ulf Wiger (TN/EAB)" <> wrote:

> Mats Cronqvist skrev:
> > Benjamin Tolputt wrote:
> >>
> >> The issue is that taking the direction you suggest *right now*
>  >> will have the effect of making Erlang slower for most tasks on
>  >> most platforms it is currently deployed on.
> >   Will it? I doubt it would make a realistic application (e.g.
>  >   a phone switch) significantly slower even on single core HW.
>  >   Perhaps you're aware of some measurement that would prove me
>  >   wrong?
> 
> The shootout could be a place to experiment with this, if we
> could have one Erlang entry that assumes SMP, and one that
> doesn't, for each benchmark (like we used to have a non-HiPE
> and a HiPE version).

Well, we seem to be heading towards different data sets
- q6600 all four cores
- q6600 just one core
with programs that use different strategies for the same algorithm
listed together (yes, I know, it's getting a bit nuanced).


> One issue is that there isn't e.g. a plists in the stdlib.
> With permission from the shootout maintainers, we could
> possibly have a patched build, with a lists module (etc.)
> defaulting to parallel solutions (i.e. spawning processes
> where sensible, and gathering results, preserving order
> where necessary.)

Yuck - possibly maybe as a very temporary thing.

Wouldn't it be better to try this with fab 24 CPU hardware?



      



More information about the erlang-questions mailing list